Obesity
Original Article
CLINICAL TRIALS AND INVESTIGATIONS
Effect of Reducing Portion Size at a Compulsory Meal
on Later Energy Intake, Gut Hormones, and Appetite in
Overweight Adults
Hannah B. Lewis1, Amy L. Ahern1, Ivonne Solis-Trapala1, Celia G. Walker1, Frank Reimann2, Fiona M. Gribble2,
and Susan A Jebb1
Objective: Larger portion sizes (PS) are associated with greater energy intake (EI), but little evidence
exists on the appetitive effects of PS reduction. This study investigated the impact of reducing breakfast
PS on subsequent EI, postprandial gastrointestinal hormone responses, and appetite ratings.
Methods: In a randomized crossover design (n 5 33 adults; mean BMI 29 kg/m2), a compulsory breakfast
was based on 25% of gender-specific estimated daily energy requirements; PS was reduced by 20%
and 40%. EI was measured at an ad libitum lunch (240 min) and snack (360 min) and by weighed diet
diaries until bed. Blood was sampled until lunch in 20 participants. Appetite ratings were measured using
visual analogue scales.
Results: EI at lunch (control: 2,930 6 203; 20% reduction: 2,853 6 198; 40% reduction: 2,911 6 179 kJ)
and over the whole day except breakfast (control: 7,374 6 361; 20% reduction: 7,566 6 468; 40% reduction: 7,413 6 417 kJ) did not differ. Postprandial PYY, GLP-1, GIP, insulin, and fullness profiles were lower
and hunger, desire to eat, and prospective consumption higher following 40% reduction compared to
control. Appetite ratings profiles, but not hormone concentrations, were associated with subsequent EI.
Conclusions: Smaller portions at breakfast led to reductions in gastrointestinal hormone secretion but
did not affect subsequent energy intake, suggesting small reductions in portion size may be a useful
strategy to constrain EI.
Obesity (2015) 23, 1362–1370. doi:10.1002/oby.21105
Introduction
Concurrent with the increasing prevalence of obesity has been the
increased mass of food consumed per eating occasion (1) and the
increased size of commercially available portions (2). Empirical evidence shows larger portion sizes (PS) lead to greater energy intake
(EI) at single meals, an effect that continues through 11 days of
manipulation (3,4). Reducing PS is a central component in weight
management advice, but experimental work to investigate whether
PS reduction leads to changes in subsequent EI is limited. Given
that there are strong homeostatic mechanisms to guard against low
energy intakes (5), energy compensation may occur in an environment where food is widely available. Gastrointestinal hormones and
the perception of appetite are important biological and psychological
appetitive mechanisms. There has been little simultaneous research
of these mechanisms and energy intake responses following modest
portion size manipulations of mixed meals.
Moreover, most research in relation to the effect of portion size
on intake has been conducted among individuals with a healthy
weight where appetite control systems may be at their most
robust. In contrast, overweight and obese individuals are poorly
studied yet represent the population most in need of interventions
to constrain energy intake. This study investigated whether reducing the PS of a compulsory meal is an effective strategy to reduce
day-long EI in overweight and obese adults. This study also investigated the impact on gastrointestinal hormones and appetite ratings as measures of biological and psychological appetite control
mechanisms.
1
Diet and Obesity Research, Medical Research Council Human Nutrition Research, Cambridge, UK. Correspondence: Hannah B. Lewis (hannah.b.lewis@
cantab.net) 2 Department of Clinical Biochemistry, Institute of Metabolic Science, University of Cambridge, UK.
Funding agencies: This study was supported by a programme grant from the UK Medical Research Council (U105960389). GLP-1 analysis was funded by Takeda
Cambridge Ltd., UK.
Disclosure: SAJ is the independent Chair of the Department of Health Responsibility Deal Food Network in England, which includes voluntary agreements with industry to
reduce the portion size of some food and drinks. No other authors declare a conflict of interest.
Author contributions: HBL and SAJ were responsible for project conception. HBL, ALA, and SAJ developed the protocol. IS-T advised on statistical analysis. HBL
conducted research, analyzed data, interpreted results, and drafted the manuscript. ALA, IS-T, CGW, FR, FMG, and SAJ contributed to the data interpretation and critical
revision of the manuscript. All authors read and approved the final manuscript.
Received: 1 October 2014; Accepted: 11 March 2015; Published online 5 June 2015. doi:10.1002/oby.21105
1362
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
www.obesityjournal.org
Original Article
Obesity
CLINICAL TRIALS AND INVESTIGATIONS
Methods
Study design
The study was a randomized crossover design involving three PS
conditions, presented to each participant at a standardized breakfast
time on separate days: a control PS; PS reduced by 20%; and PS
reduced by 40%. The control provided 25% of estimated daily
energy requirements for the intended average study participant
according to gender (6), the energy content of a representative main
meal. The test breakfast was consumed in its entirety. Participants
were blinded to the specific aims of the study and foods prepared to
draw attention away from the intervention. For each individual,
study visits were conducted >1-week apart, on the same weekday
and outside the luteal phase of the menstrual cycle for females.
Participants
Healthy, 18-60 year men and women, with a BMI 25 and <35
kg/m2 were recruited. Participants were excluded for disordered eating [Eating Attitudes Test (EAT-26) score 11 (7,8)], depressive
symptoms [Zung Depression Scale score 70 (9)], smoking, excessive habitual alcohol intake (women > 14 units/week, men >21
units/week), weight loss/gain within last 3 months (>4.5 kg) or
actively trying to lose/gain weight, medical conditions/medications
potentially affecting appetite, inflammatory conditions, diabetes or
fasting plasma glucose 7 mmol/L, pregnancy, breastfeeding or
planning pregnancy, extremely high levels of exercise [moderate/
vigorous activity >420 min/week assessed with International Physical Activity Questionnaire (IPAQ) (10)], unable to eat test foods,
and not regularly consuming breakfast (breakfast 3/week).
A sample size of 33 was recruited to give 83% power to detect a
minimum difference of 500 kJ lunch EI between any pair of experimental conditions assuming an SD of 950 kJ (3,11,12). Biochemical
measures were collected in a subgroup of 20 participants.
blood samples were collected at fasting and 30, 60, 120, 180, and
240 min for the analysis of peptide tyrosine tyrosine 3-36 (PYY336), total glucagon-like peptide-1 (GLP-1), total glucose-dependent
insulinotropic peptide (GIP), glucose, and insulin (Figure 1).
At the end of the study participants were fully debriefed on the
study aims and asked to verbally report if they had noticed the portion size manipulation between visits. They were reimbursed for
travel expenses and given an honorarium. Ethical approval for the
study was obtained from Cambridgeshire 2 Research Ethics Committee in November 2010 (Ref: 10/H0308/99) and participants gave
informed written consent. The study was conducted at Medical
Research Council Human Nutrition Research (MRC HNR) between
January 2011 and September 2012.
Study foods
The breakfast consisted of a wheat-based breakfast cereal with semiskimmed milk, scrambled egg, ham, brown toast and butter, and
orange juice. For men and women respectively, the control breakfast
provided 3,310 and 2,540 kJ; the 20% reduction breakfast provided
2,650 kJ and 2,030 kJ; the 40% reduction breakfast provided 1,990
and 1,520 kJ. The ad libitum lunch consisted of a single course
amorphous meal of pasta, minced beef, tomato sauce, mixed vegetables, and grated cheese. The lunch provided 8,275 kJ (men) or 6,350
kJ (women). The study breakfast and lunch provided 35% energy
from fat, 18% from protein and 47% from carbohydrates (14). The
ad libitum snack consisted of ten digestive biscuits (plain cookies)
on a plate (3,250 kJ).
The completed diet diaries and weighed consumption at lunch and
snack for each study day were used to calculate energy intake using
MRC HNR’s dietary assessment system (15).
Questionnaires
Recruitment and screening
Participants were recruited from the community, for a study investigating the “relationship between diet and metabolism.” Height,
weight, waist circumference, body composition (Tanita body composition analyzer BC-418MA), and resting metabolic rate (RMR; IS
Gem204 with GEMNutrition 2008.4 software) were measured. Participants completed the EAT-26, Zung depression scale, IPAQ and
the Three Factor Eating Questionnaire measuring dietary restraint,
disinhibition, and hunger traits (13), and fasting plasma glucose was
assessed. Participants were asked to maintain their usual exercise
and dietary habits during the study.
The appetite VAS questionnaires rated hunger, fullness, desire to eat
and prospective consumption, and also included five distractor questions on mood. The palatability questionnaire used VAS to rate palatability parameters, as distractor questions, and portion size. The
VAS questionnaires asked participants to mark a horizontal line
measuring 100 mm with the ends labelled with the extremes of each
sensation (e.g. “Not at all” and “Extremely”). The portion size question was “How large is the size of the portion?” anchored with
“Extremely small” and “Extremely large.” The distance from the
left end to the mark was measured to the nearest millimetre.
Analytic methods
Study visits
Participants fasted overnight (11 h prior to each visit) and were
asked to refrain from alcohol and avoid strenuous exercise the 24 h
before each study day. Provision of the test breakfast marked time
zero. Subsequent EI was measured by premeal and postmeal weighing of an ad libitum lunch (240 min) and afternoon snack (360
min), plus a weighed diet diary to record the remainder of the day’s
intake. Visual analogue scale (VAS) questionnaires rating palatability and portion size were given after the first bite of food at breakfast and lunch. Appetite ratings were measured using VAS questionnaires at 30-min intervals until lunch, then immediately after and at
300 and 360 min, then hourly. In a subgroup of 20 participants,
www.obesityjournal.org
Blood samples were separated on collection and plasma stored at
280 C. Plasma collected on EDTA and treated with dipeptidyl
peptidase-IV (DPP-IV) inhibitor immediately on collection (10 lL
DPP-IV inhibitor/ml of blood) was analyzed for PYY3-36 by radioR , MA) (interassay CVs: 15% at 84 pg/mL;
immunoassay (MilliporeV
7% at 217 pg/mL), at University College Hospital, London; total
GLP-1 using an electrochemical luminescence immunoassay kit on
R multiarray assay platform (MD, USA)
the MesoScale DiscoveryV
(CVs: 16.4%, 11.9% and 11.6% at 5.4, 29, and 83 pg/mL, respectively), at Core Biochemical Assay Laboratory (CBAL), Cambridge;
and total GIP using an enzyme-linked immunosorbent assay (MilliR , MA) (CVs: 6.1, 3.3, 2.3, and 1.8% at 26, 50, 134, and 166
poreV
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
1363
Obesity
Portion Size Reduction and Appetite Control Lewis et al.
Figure 1 Overview of the time points for meals and measurements taken during a study day. GIP: glucosedependent insulinotropic peptide; GLP-1: glucagon-like peptide 1; MRC HNR: Medical Research Council Human
Nutrition Research; PYY: peptide tyrosine tyrosine; VAS: visual analogue scales.
pg/mL, respectively), at Cambridge Institute for Medical Research.
Plasma collected on fluoride oxalate was analyzed for glucose using
R clinical chemistry system (Siemens, NE) (CVs: 1.69,
a DimensionV
2.23, and 2.56% at 6.23, 3.09, and 18.88 mmol/L respectively), at
MRC HNR. Plasma collected on lithium heparin was analyzed for
R automatic immunoassay analyzer
insulin on a 1235 AutoDELFIAV
using a two-step time resolved fluorometric assay (Perkin Elmer
Life Sciences, Wallac Oy, Turku, Finland) (CVs: 3.1, 2.1, 1.9, and
2.0% at 29, 79.4, 277, and 705 pmol/L, respectively) at CBAL,
Cambridge.
Statistical analysis
Mixed effects models for continuous responses (16) were used for
analysis, which extend standard linear regression to account for correlation between repeated observations within individuals through
random effects. EI and PS ratings at breakfast were modeled with
PS condition as the explanatory variable, controlling for gender and
BMI. Dietary restraint, disinhibition and hunger, were tested for
inclusion as covariates, but were omitted for no effects on the associations of interest.
The effect of PS condition on biochemical measures and appetite
ratings was assessed by the interaction between condition and time,
which estimated differences at each time point. Area under the curve
(AUC) was calculated using the trapezoidal rule for the time periods
of fasting to the prelunch time point for biochemical measures and
appetite ratings, and over the whole day for appetite ratings. Models
of whole-day appetite ratings AUC included PS condition as the
1364
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
explanatory variable, controlling for the time over which appetite
ratings were made.
Models predicting EI at lunch included explanatory variables of
either the prelunch or AUC for each biochemical measure or appetite rating, and controlled for condition, gender and BMI. Similar
models assessed the relationship between the whole-day AUC of
appetite rating with whole day EI (except breakfast), also controlling
for time over which appetite ratings were made.
To examine the relationship between biochemical measures and
appetite ratings, appetite ratings were modeled separately with each
biochemical measure as the explanatory variable. Time, a quadratic
term for time, condition, gender, and BMI were covariates.
Potential carry-over and sequence effects, gender, BMI and age,
unless specified above as included a priori, were omitted as covariates as there were no effects on the associations of interest. To
account for correlation induced by multiple observations/individuals
(three visits), a random intercept was incorporated into the models.
The models for biochemical and appetite ratings profiles as outcomes had two levels of clustering due to repeated sampling time
points and the crossover design. Therefore, a random intercept and
slope for time were added to model within-individual variation.
Models were fitted using maximum likelihood estimation and likelihood ratio tests were used for model comparison. Plots of residuals
were used to check the goodness of fit for each outcome. Insulin
and GIP data were transformed (natural logarithm and square root
respectively) for analyses, for a symmetrical distribution. All analyR 12.0
ses used STATAV
software (StataCorp, TX). Statistical
www.obesityjournal.org
Original Article
Obesity
CLINICAL TRIALS AND INVESTIGATIONS
TABLE 1 Participant characteristics
Number of men/women
Height (m)
Weight (kg)
BMI (kg/m2)
Age (years)
Dietary restraint
Disinhibition
Hunger trait
RMR (kJ/day)
Fasting glucose (mmol/L)
Body fat (%)
Vigorous physical activity (min per week)
Moderate physical activity (min per week)
Walking (min per week)
All participants
(n 5 33)
Blood sample
subgroup (n 5 20)
Nonblood subgroup
(n 5 13)
15/18
1.69 6 0.01
83.8 6 1.5
29.0 6 0.4
42.5 6 2.0
7.2 6 0.7
6.7 6 0.6
6.3 6 0.7
6594 6 160
4.8 6 0.1
32.8 6 1.5
65 6 13
142 6 21
254 6 30
9/11
1.69 6 0.01
82.9 6 2.1
29.0 6 0.5
40.8 6 2.5
6.5 6 0.9
6.5 6 0.7
6.2 6 0.8
6704 6 224
4.7 6 0.1
31.9 6 1.8
55 6 14
173 6 29
270 6 37
7/6
1.71 6 0.03
85.3 6 2.0
29.2 6 0.8
45 6 3.4
8.2 6 1.1
6.9 6 1.1
6.5 6 1.1
6425 6 220
4.9 6 0.1
34.2 6 2.6
80 6 24
94 6 26
231 6 53
Mean 6 SEM.
BMI: body mass index; RMR: resting metabolic rate.
significance was set at P < 0.05. Data are presented as mean 6 SEM
unless indicated otherwise.
Results
Participant characteristics
The characteristics of the study participants are shown in Table 1.
Energy intake
EI was not different between conditions at lunch (control:
2,930 6 203; 20% reduction condition: 2,853 6 198; 40% reduction
condition: 2,911 6 179 kJ) or over the whole day except breakfast
(control: 7,374 6 361; 20% reduction condition: 7,566 6 468; 40%
reduction condition: 7,413 6 417 kJ) (Figure 2). Daily EI including
breakfast was 10,287 6 395 kJ, 9,897 6 491 kJ, and 9,161 6 437 kJ
in control, 20% reduction and 40% reduction conditions
respectively.
Biochemical measures
Figure 3 shows the postprandial profiles for each of the gastrointestinal hormones. Compared to control, there was a reduction in PYY
in 40% reduction condition at 120 and 240 min (P < 0.05), but no
condition-time interaction for 40% reduction compared to 20%
reduction condition (P > 0.076), or 20% reduction compared to control (P > 0.42). Compared to control, GLP-1 was lower in 40%
reduction condition at all postprandial time points (P < 0.05). GLP-1
was also lower in 40% reduction condition compared to 20% reduction at 180 min (P 5 0.038), but there was no condition-time interaction for 20% reduction condition compared to control (P > 0.056).
GIP was lower in 40% reduction condition compared to control at
all postprandial time points (P < 0.001), and compared to 20%
reduction condition at 30, 120, 180, and 240 min (P < 0.05). GIP
www.obesityjournal.org
Figure 2 Mean (6 SEM) energy intake (A) at lunch (control vs. 20% reduction,
b 5 276.6, P 5 0.429; 20% reduction vs. 40% reduction, b 5 58.2, P 5 0.547; control vs. 40% reduction, b 5 218.3, P 5 0.850) and (B) over the whole day except
breakfast (control vs. 20% reduction, b 5 192.3, P 5 0.555; 20% reduction vs. 40%
reduction, b 5 2152.8, P 5 0.639; control vs. 40% reduction, b 5 39.5, P 5 0.904)
did not differ between conditions.
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
1365
Obesity
Portion Size Reduction and Appetite Control Lewis et al.
similarly in 20% reduction condition compared to control. Desire to
eat ratings were generally greater in 40% reduction condition compared to control (P < 0.023) and compared to 20% reduction condition (P < 0.037). There was no condition-time interaction for 20%
reduction condition compared to control (P > 0.223). Prospective
consumption was greater in 40% reduction condition compared to
control (P < 0.011) but there were fewer differences when compared
to 20% reduction condition. There was no condition-time interaction
for 20% reduction condition compared to control (P > 0.068).
AUCs over the whole day for hunger, desire to eat and prospective
consumption were greater in 40% reduction condition compared to
control, and smaller for fullness (hunger b 5 2423.9, P 5 0.025; fullness b 5 24857.9, P 5 0.001; desire to eat b 5 3832.5, P 5 0.001;
prospective consumption b 5 3427.9, P 5 0.001). AUC for prospective consumption ratings was greater in 20% reduction condition
compared to control (b 5 2284.1, P 5 0.025), but AUC for hunger
(P 5 0.232), fullness (P 5 0.136), and desire to eat (P 5 0.118) did
not differ. There were no differences in hunger or fullness when
comparing 20 and 40% reduction conditions (data not shown).
Predictors of energy intake
Most of the biochemical measures did not predict EI at lunch
(P > 0.137) (Table 2). However, AUC (P 5 0.032) and prelunch
(P 5 0.049) measures of PYY were positively associated with EI at
lunch. AUCs and prelunch measures of hunger, desire to eat and
prospective consumption were positively associated with lunch EI
(P < 0.02). Prelunch fullness was negatively associated with lunch
Figure 3 Postprandial response (mean 6 SEM) of (A) plasma PYY3-36, (B) plasma
total GLP-1, and (C) plasma total GIP, according to condition. “a” indicates the
mean of the condition is significantly different from the mean of the control condition at that time point. “b” indicates the mean is significantly different from the
mean of the 20% reduction condition at that time point (mixed effects models):
P < 0.05. Addition of * to the letter indicates P < 0.01 and ** indicates P < 0.001.
was lower at 120, 180, and 240 min in 20% reduction condition
compared to control (P < 0.05).
Glucose and insulin profiles are shown in Figure 4. There was no
condition-time interaction for glucose (P > 0.2 for all comparisons).
There was a condition-time interaction such that insulin was less in
40% reduction condition compared to control and 20% reduction
condition at 120 and 180 (P < 0.05), and compared to control at 240
min (P < 0.01). There was no condition-time interaction for 20%
reduction condition compared to control (P > 0.083).
Appetite ratings
Figure 5 shows the appetite ratings. Hunger was greater after breakfast in 40% reduction condition compared to control (P < 0.006) and
to 20% reduction condition (P < 0.021). There was no conditiontime interaction for 20% reduction condition compared to control
(P > 0.291). There were some differences in postprandial fullness
between conditions whereby fullness was lower in 40% reduction
condition compared to control (P < 0.019). However there were
fewer differences when compared to 20% reduction condition, and
1366
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
Figure 4 Postprandial response of (A) plasma glucose (mean 6 SEM) and (B)
plasma insulin (geometric mean 6 95% confidence intervals), according to condition. “a” indicates the mean of the condition is significantly different from the mean
of the control condition at that time point. “b” indicates the mean is significantly different from the mean of the 20% reduction condition at that time point (mixed
effects models): P < 0.05. Addition of * to the letter indicates P < 0.01 and ** indicates P < 0.001.
www.obesityjournal.org
Original Article
Obesity
CLINICAL TRIALS AND INVESTIGATIONS
Figure 5 Postprandial ratings (mean 6 SEM) for (A) hunger, (B) fullness, (C) desire to eat, and (D) prospective consumption, according to condition. “a”
indicates the mean of the condition is significantly different from the mean of the control condition at that time point. “b” indicates the mean is significantly different from the mean of the 20% reduction condition at that time point (mixed effects models): P < 0.05. Addition of * to the letter indicates
P < 0.01 and ** indicates P < 0.001.
EI (P < 0.002), but fullness AUC was not (P 5 0.085). AUCs for
hunger, desire to eat and prospective consumption, but not fullness
(P 5 0.469), were positively associated with EI over the day
(P < 0.026).
Biochemical measures and appetite ratings
associations
GLP-1, GIP, glucose and insulin were negatively associated with
hunger, desire to eat, and prospective consumption, and positively
associated with fullness (P < 0.012). PYY was not associated with
any of the appetite ratings (P > 0.068) (Table 3).
Perceived portion size
The ratings of portion size at breakfast were different between conditions. Portion size was perceived to be smaller in 40% reduction
condition compared to both control (b 5 215.6, P < 0.001) and 20%
reduction condition (b 5 210.8, P < 0.001), and portion size was
perceived to be smaller in 20% reduction condition compared to
control (b 5 24.8, P 5 0.049) (portion size rating: control
53 6 2 mm; 20% reduction 48 6 3 mm; 40% reduction 37 6 3 mm).
However at debriefing only two participants specifically reported
noticing the change in PS at breakfast. None of the participants
were concerned about the blinding of the aims of the study and all
consented to data inclusion.
www.obesityjournal.org
Discussion
Reducing PS at a single meal altered biological markers of appetite
and appetite ratings, but there was no energy compensation later in the
day. EIs at lunch were strikingly consistent in this standardized laboratory setting. These findings indicate reducing PS of a prepared meal
could lead to a net reduction in daily EI. However, the effect on gastrointestinal hormones and appetite ratings, particularly after the 40%
reduction in PS, questions the sustainability of this strategy to constrain EI.
There were very few differences in the profiles for PYY and GLP-1
between control and 20% reduction conditions. Moreover, there were
few differences in the profiles when comparing 20 and 40% reduction
conditions suggesting that the responses in these biochemical measures
may not be sensitive to the smaller change in PS (660 kJ men and 510
kJ women). Indeed, all previous studies where a reduction in energy
load has led to attenuated PYY (17,18), GLP-1 (19,20), or insulin
(21,22) profiles, used energy changes between 920 and 2,096 kJ. However, the present study showed distinct differences between all conditions in the postprandial profiles for GIP showing that it is sensitive to
energy changes in a clear dose response manner, reflecting its important role as an incretin hormone for the regulation of insulin secretion.
Interestingly, the ratings of perceived PS of the breakfast were different between conditions, although at debriefing most participants
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
1367
Obesity
Portion Size Reduction and Appetite Control Lewis et al.
TABLE 2 Estimated regression coefficients to measure associations between biochemical measures and appetite ratings
(predictor variables) with energy intake at lunch and over the whole day apart from breakfast (outcome variables), from
mixed effects models
AUC as predictor
Regression
coefficient (SE)
Predictor of lunch EI
Biochemical measure
PYY
GLP-1
GIP
Glucose
Insulin
Appetite rating
Hunger
Fullness
Desire to eat
Prospective consumption
Predictor of whole day EI
Regression
coefficient (SE)
P value
P value
0.029
0.019
2.666
20.916
2197.0
(0.014)
(0.071)
(4.446)
(0.710)
(259.8)
0.032
0.790
0.549
0.197
0.448
4.442
15.95
39.08
2365.5
2157.0
(2.257)
(11.17)
(33.06)
(245.8)
(168.7)
0.049
0.154
0.237
0.137
0.352
0.091
20.029
0.087
0.100
(0.022)
(0.017)
(0.018)
(0.022)
<0.001
0.085
<0.001
<0.001
11.96
210.43
8.788
19.21
(3.934)
(3.389)
(3.783)
(4.384)
0.002
0.002
0.020
<0.001
Regression
coefficient (SE)
AUC appetite rating
Hunger
Fullness
Desire to eat
Prospective consumption
Prelunch measure as predictor
0.057
20.016
0.057
0.068
P value
(0.025)
(0.021)
(0.023)
(0.025)
0.026
0.469
0.013
0.007
AUC: area under the curve. EI: energy intake. SE: standard error.
Area under the curve was calculated for between the fasting and prelunch time points for predicting energy intake at lunch. Area under the curve for the whole day was
calculated for predicting energy intake over the whole day apart from breakfast. Each predictor was analyzed in a separate mixed effects model.
Values are given to four significant figures. Those in bold are significant.
reported not noticing the meal manipulation. The effect size for the
difference between perceived PS ratings was considerably smaller
when comparing control versus 20% reduction than 20% versus
40% reduction conditions (b 5 24.8; b 5 210.8), although the abso-
lute difference in energy was the same. This difference is likely due
to either the relative difference between PS being different (20% for
control-20% reduction comparison, and 25% for 20-40% reduction
comparison), or due to the Weber-Fechner law, whereby the ability to
TABLE 3 Estimated regression coefficients to measure associations between biochemical measures (predictor variables) and
appetite ratings (outcome variables) from baseline to the prelunch time point, from mixed effects models
Appetite rating
Hunger
Biochemical
measure
PYY
GLP-1
GIP
Glucose
Insulin
Regression
coefficient (SE)
20.032
20.494
23.271
26.650
214.07
(0. 038)
(0.172)
(0.373)
(1.058)
(1.227)
Fullness
P value
Regression
coefficient (SE)
0.409
0.004
<0.001
<0.001
<0.001
0.041
0.631
3.357
6.058
14.33
(0.041)
(0.186)
(0.416)
(1.186)
(1.391)
Desire to eat
P value
0.315
0.001
<0.001
<0.001
<0.001
Regression
coefficient (SE)
20.018
20.442
23.143
25.493
213.63
(0.040)
(0.176)
(0.379)
(1.087)
(1.250)
P value
0.650
0.012
<0.001
<0.001
<0.001
Prospective
consumption
Regression
coefficient (SE)
20.028
20.421
22.629
24.396
211.86
(0.031)
(0.138)
(0.305)
(0.884)
(0.990)
P value
0.366
0.002
<0.001
<0.001
<0.001
SE: standard error.
Each predictor was analyzed in a separate mixed effects model.
Values are given to 4 significant figures. Those in bold are significant.
1368
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
www.obesityjournal.org
Original Article
Obesity
CLINICAL TRIALS AND INVESTIGATIONS
perceive stimulus change is proportional to the logarithm of the magnitude of the stimulus (23). Thus, as the reference portion size in the
first comparison (control versus 20% reduction) was larger than the
second (20% versus 40% reduction), the change in PS detectable for
the first pairing would have been larger than the second. It is possible
that the perception of how much is provided, and thus consumed,
could affect appetite ratings. The smaller effect size of perceived PS
between control and 20% reduction conditions could in part account
for fewer differences in appetite ratings between these conditions.
Postprandial biochemical responses were poor predictors of subsequent EI, consistent with much of the existing evidence (24-26).
However, appetite ratings tended to predict EI at lunch and the rest
of the day. This is in agreement with some (12,27-29), but not all
(30,31), previous studies. The mixed evidence likely reflects the
subjective nature of the perception of appetite which leads to measurement variability, but differences are more easily detected in
crossover than parallel design studies (32). Although associations
between appetite ratings and EI in the present study were highly significant, the effect sizes were small. This, coupled with relatively
small differences in postprandial appetite ratings responses to the
manipulated meal, could in part explain the lack of compensation
for the changes in energy. In contrast with the known function of
PYY, where exogenous administration reduces EI (17,33,34), there
was a small but significant positive effect of AUC and prelunch
PYY on subsequent EI. However, the effect decreased after adjustment for additional participant characteristics, indicating it may be
confounded by other factors. Thus there is uncertainty about these
present findings relating to PYY. In contrast to the clear exogenous
effect, endogenous postprandial responses in PYY were not associated with subsequent EI (12,25,35), possibly as exogenous PYY
tends to be used at supra-physiological levels (12).
GLP-1, glucose and insulin were positively related to fullness and
negatively related to hunger, desire to eat and prospective consumption consistent with previous research (22,24,36-38), indicating that
these biochemical measures are likely to play roles in the perception
of appetite sensations. However, some studies have found no relationship, or mixed results, between glucose or insulin and appetite
ratings (24,29), possibly because they have reported correlations
between the mean AUC or peak values rather than examining
within-person relationships. Previous findings with respect to the
relationship between postprandial PYY and appetite are mixed,
including positive associations between PYY and perceived fullness
(36,39), while others, consistent with the present findings, have
found no associations (12,38), or associations in lean but not obese
participants (35). Thus, the robustness of the association of
endogenous PYY with appetite ratings is questionable. It is unclear
whether GIP plays a role in influencing appetite and EI (40), however the present findings showed GIP was associated with appetite
ratings. The distinct similarity between GIP and appetite ratings
profiles may have led to these associations, but causality cannot be
assumed. The lack of association between GIP and subsequent
lunch EI is in agreement with the perspective that GIP does not
influence EI.
The present findings support the concept that reducing the PS of
prepared meals or unit foods could constrain EI and contribute to
prevention of weight gain. However as weight control advice is
inherently overt, it is important to establish whether similar effects
are seen when participants are aware of the reduction in PS.
www.obesityjournal.org
There are several limitations to this study. It was conducted in a laboratory setting and, although the specific hypothesis was concealed,
participants were aware of their eating behavior being observed. The
frequency and type of food provided at lunch was fixed, thus only
the amount could vary potentially limiting compensation by removing some of the environmental cues that are profuse in a free-living
environment and can influence EI. This setting also prevented any
self-initiated eating episodes between breakfast and lunch. Some of
the appetite and hormone profiles suggest effects of PS reduction
may have diminished over time and compensation might be seen in
a free-living environment during this period. The use of a different
control PS (in terms of percentage of estimated daily energy requirements) may alter the relative effects of a percentage reduction. The
overweight and obese sample may limit comparison with previous
work, however it was most appropriate to study portion size reduction in this group as those with an elevated BMI are most in need of
interventions to constrain energy intake, and are to date an understudied population in this field. The study was conducted over a single day and it is possible that a longer period of consuming portions
set to provide energy below requirements could lead to adaptation
and energy compensation. Future studies should attempt to examine
PS reduction in a more realistic setting and with prolonged exposure
to smaller portions.
Conclusion
Compulsory reductions in PS at breakfast did not lead to differences
in subsequent energy intake over the day, despite changes in biological and psychological measures that tend to favor energy compensation. Although the effect size was small, if sustained this will be of
public health benefit in constraining energy intake. O
Acknowledgments
We thank the volunteers who participated and the research assistants/students who assisted with data collection. Thank you to
Johannes Grosse for discussion and enabling GLP-1 analysis. Thank
you to JJ Laboratory Services, London, for analysis of PYY; Core
Biochemical Assay Laboratory, Cambridge, for analysis of insulin
and GLP-1; MRC HNR for analysis of glucose and diet diary coding; and Biochemistry and Immunology Department at Addenbrooke’s Hospital, Cambridge for screening fasting plasma glucose
analysis.
C 2015 The Obesity Society
V
References
1. Nielsen SJ, Popkin BM. Patterns and trends in food portion sizes, 1977-1998.
JAMA 2003;289:450-453.
2. Young LR, Nestle M. The contribution of expanding portion sizes to the US obesity
epidemic. Am J Public Health 2002;92:246-249.
3. Rolls BJ, Morris EL, Roe LS. Portion size of food affects energy intake in normalweight and overweight men and women. Am J Clin Nutr 2002;76:1207-1213.
4. Rolls BJ, Roe LS, Meengs JS. The effect of large portion sizes on energy intake is
sustained for 11 days. Obesity 2007;15:1535-1543.
5. Prentice A, Jebb S. Energy intake/physical activity interactions in the homeostasis
of body weight regulation. Nutr Rev 2004;62:S98-S104.
6. Scientific Advisory Committee on Nutrition (2011). Dietary Reference Values for
Energy. London: TSO. [www document] Available at: URL: http://www.sacn.gov.uk/
reports_position_statements/reports/sacn_dietary_reference_values_for_energy.html
7. Garner DM, Garfinkel PE. The eating attitudes test: an index of the symptoms of
anorexia nervosa. Psychol Med 1979;9:273-279.
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
1369
Obesity
Portion Size Reduction and Appetite Control Lewis et al.
8. Orbitello B, Ciano R, Corsaro M, et al. The EAT-26 as screening instrument for
clinical nutrition unit attenders. Int J Obes 2006;30:977-981.
9. Zung WWK. A self-rating depression scale. Arch Gen Psychiatry 1965;12:63-70.
10. The IPAQ Group. International Physical Activity Questionnaire. [www document]
Available at: URL: https://sites.google.com/site/theipaq/
11. Kral TVE, Roe LS, Rolls BJ. Combined effects of energy density and portion size
on energy intake in women. Am J Clin Nutr 2004;79:962-968.
12. Doucet E, Laviolette M, Imbeault P, Strychar I, Rabasa-Lhoret R, Prud’homme D.
Total peptide YY is a correlate of postprandial energy expenditure but not of
appetite or energy intake in healthy women. Metabolism 2008;57:1458-1464.
13. Stunkard AJ, Messick S. The three-factor eating questionnaire to measure dietary
restraint, disinhibition and hunger. J Psychosom Res 1985;29:71-83.
14. Food Standards Agency and Department of Health (2010). National Diet and
Nutrition Survey: Headline results from Year 1 of the Rolling Programme (2008/
2009). [www document] Available at: URL: http://www.food.gov.uk/multimedia/
pdfs/publication/ndnsreport0809year1results.pdf
15. Fitt E, Cole D, Ziauddeen N, et al. DINO (Diet In Nutrients Out)—an integrated
dietary assessment system. Public Health Nutrition 2014;27:1-8.
16. McCulloch C, Searle S. Generalized Linear and Mixed Models. New York: Wiley;
2000.
17. le Roux CW, Batterham RL, Aylwin SJB, et al. Attenuated peptide YY release in
obese subjects is associated with reduced satiety. Endocrinology 2006;147:3-8.
18. Martins C, Robertson MD, Morgan LM. Impact of restraint and disinhibition
on PYY plasma levels and subjective feelings of appetite. Appetite 2010;55:208213.
19. Vilsbøll T, Krarup T, Sonne J, et al. Incretin secretion in relation to meal size and
body weight in healthy subjects and people with type 1 and type 2 diabetes
mellitus. J Clin Endocrinol Metab 2003;88:2706-2713.
20. Rijkelijkhuizen JM, McQuarrie K, Girman CJ, et al. Effects of meal size and
composition on incretin, a-cell, and b-cell responses. Metabolism 2010;59:502511.
21. Borer KT, Wuorinen E, Ku K, Burant C. Appetite responds to changes in meal
content, whereas ghrelin, leptin, and insulin track changes in energy availability.
J Clin Endocrinol Metab 2009;94:2290-2298.
22. Blom WAM, Stafleu A, de Graaf C, Kok FJ, Schaafsma G, Hendriks HFJ. Ghrelin
response to carbohydrate-enriched breakfast is related to insulin. Am J Clin Nutr
2005;81:367-375.
23. Colman AM. Oxford Dictionary of Psychology, 3rd ed. Oxford: Oxford University
Press; 2009.
24. Flint A, Moller BK, Raben A, et al. Glycemic and insulinemic responses as
determinants of appetite in humans. Am J Clin Nutr 2006;84:1365-1373.
25. Willbond SM, Doucet E. Individually timing high-protein preloads has no effect on
daily energy intake, peptide YY and glucagon-like peptide-1. Eur J Clin Nutr 2011;
65:55-62.
1370
Obesity | VOLUME 23 | NUMBER 7 | JULY 2015
26. Verdich C, Toubro S, Buemann B, Lysgård Madsen J, Juul Holst J, Astrup A. The
role of postprandial releases of insulin and incretin hormones in meal-induced
satiety—effect of obesity and weight reduction. Int J Obes 2001;25:1206-1214.
27. Drapeau V, King N, Hetherington M, Doucet E, Blundell J, Tremblay A. Appetite
sensations and satiety quotient: predictors of energy intake and weight loss. Appetite
2007;48:159-166.
28. Lemmens SG, Martens EA, Born JM, Martens MJ, Westerterp-Plantenga MS.
Staggered meal consumption facilitates appetite control without affecting
postprandial energy intake. J Nutr 2011;141:482-488.
29. Holt SHA, Miller JCB, Petocz P. Interrelationships among postprandial satiety,
glucose and insulin responses and changes in subsequent food intake. EurJ Clin
Nutr 1996;50:788-797.
30. Gray RW, French SJ, Robinson TM, Yeomans MR. Dissociation of the effects of
preload volume and energy content on subjective appetite and food intake. Physiol
Behav 2002;76:57-64.
31. De Graaf C, Hulshof T. Effects of weight and energy content of preloads on
subsequent appetite and food intake. Appetite 1996;26:139-151.
32. Stubbs RJ, Hughes DA, Johnstone AM, et al. The use of visual analogue scales to
assess motivation to eat in human subjects: a review of their reliability and validity
with an evaluation of new hand-held computerized systems for temporal tracking of
appetite ratings. Br J Nutr 2000;84:405-415.
33. Batterham RL, Cohen MA, Ellis SM, et al. Inhibition of food intake in obese
subjects by peptide YY3-36. N Engl J Med 2003;349:941-948.
34. Sloth B, Holst JJ, Flint A, Gregersen NT, Astrup A. Effects of PYY1-36 and PYY336 on appetite, energy intake, energy expenditure, glucose and fat metabolism in
obese and lean subjects. Am J Physiol Endoc M 2007;292:E1062-E1068.
35. Brennan IM, Luscombe-Marsh ND, Seimon RV, et al. Effects of fat, protein, and
carbohydrate and protein load on appetite, plasma cholecystokinin, peptide YY, and
ghrelin, and energy intake in lean and obese men. Am J Physiol Gastr L 2012;303:
G129-G140.
36. Lemmens SG, Martens EA, Kester AD, Westerterp-Plantenga MS. Changes in gut
hormone and glucose concentrations in relation to hunger and fullness. Am J Clin
Nutr 2011;94:717-725.
37. Blundell JE, Levin F, King NA, et al. Overconsumption and obesity: peptides and
susceptibility to weight gain. Regul Pept 2008;149:32-38.
38. Gibbons C, Caudwell P, Finlayson G, et al. Comparison of postprandial profiles of
ghrelin, active GLP-1, and total PYY to meals varying in fat and carbohydrate and
their association with hunger and the phases of satiety. J Clin Endocrinol Metab
2013;98:E847-E855.
39. Guo Y, Ma LJ, Enriori PJ, et al. Physiological evidence for the involvement of
peptide YY in the regulation of energy homeostasis in humans. Obesity 2006;14:
1562-1570.
40. Paschetta E, Hvalryg M, Musso G. Glucose-dependent insulinotropic polypeptide:
from pathophysiology to therapeutic opportunities in obesity-associated disorders.
Obes Rev 2011;12:813-828.
www.obesityjournal.org